Inflation Expectations vs Actual Inflation: The Anchor Premium
Actual inflation is the system's temperature; expectations are its thermostat — and every long-duration valuation embeds a premium for the assumption that the thermostat stays fixed.
Most investors file inflation expectations data as a second-class citizen: a survey-based forecast of the real thing, useful only insofar as it predicts the prints that matter. That taxonomy has the relationship backwards. Expectations are not a forecast of the inflation process — they are an input to it, wired directly into the wage rounds and price schedules that generate future inflation. Actual inflation tells you the system's temperature. Expectations are the thermostat setting. And a central bank, whatever its statements target, spends most of its institutional energy defending the thermostat, because the temperature takes care of itself as long as the thermostat holds.
The framing this piece exists to install: the anchor premium. When long-horizon inflation expectations are anchored — when households, firms, and wage bargainers behave as though inflation will return to target regardless of what it is doing now — inflation shocks become self-limiting, policy responses can stay measured, and the volatility of the entire nominal system is compressed. Every long-duration asset price embeds this state of affairs as an assumption: the multiple on a growth equity, the term premium in a bond, the viability of the stock-bond hedge all quietly presume the anchor holds. That presumption is an asset — arguably the central bank's principal balance-sheet asset, though it appears on no balance sheet — and like most assets that appear on no balance sheet, it is priced as if permanent and repriced catastrophically when questioned.
The mechanism: the forecast that causes the outcome
Inflation expectations are unusual among macro variables in that believing them helps make them true. The channels are concrete. Wage bargainers negotiate nominal raises against the inflation they expect over the contract's life; firms setting prices for the year ahead build in the cost inflation they anticipate; households facing an expected erosion of money's value accelerate purchases and demand compensation in wages. Each channel converts expected inflation into actual nominal behaviour, and the actual behaviour is the inflation. This is why the distinction between expectations and actuals is not the distinction between forecast and reality — it is closer to the distinction between cause and effect, running through a lag. A price shock that leaves expectations untouched dissipates: wage rounds look through it, price schedules revert, and the shock exits the data on the base-effect timetable. The same shock, once absorbed into expectations, is no longer a shock — it is the new setting on the thermostat, and the system will now generate it endogenously.
Measuring the thermostat is its own discipline, because the three available windows disagree by construction. Household surveys are salience-driven — they track fuel and grocery prices far more than the basket, and they run persistently above measured inflation — but they are the expectations that actually walk into wage negotiations, which makes their changes informative even when their levels are biased. Professional forecasts hug the target by occupational habit and carry little independent information. Market-based breakevens reprice in real time, but a breakeven move bundles expected inflation with an inflation risk premium and a liquidity component, so reading it as pure expectations overstates the signal. The structure to hold: short-horizon measures of every kind mostly track current commodity and headline conditions, while long-horizon measures — distant-forward breakevens, long-run survey responses — are the anchoring gauges. The thermostat is the long end. The short end is mostly a thermometer wearing the wrong label.
Why the anchor dominates the actuals: three layers of why
The illustration: one anchor lost, one anchor held
The 1970s are the canonical record of what de-anchoring costs, stated as established history. Across that decade, successive shocks landed on an expectations structure that progressively gave way: households and wage bargainers stopped believing in any particular inflation level, indexation spread as self-defence, and each shock's inflation became the floor for the next round's expectations. By the end of the decade the thermostat itself had failed — inflation was being generated by the expectation of inflation. Restoring the anchor required policy restrictive enough to force a deep recession, and the credibility purchased at that price became the foundation asset of the following decades: the long disinflation, the compression of term premia, and the era in which bonds reliably hedged equities all trace back to the re-anchoring.
The post-pandemic episode is the counter-illustration, and the contrast is the entire lesson. Actual inflation overshot the target massively — by more, at peak, than during several 1970s years — yet long-horizon expectations measures moved comparatively little throughout. Wage growth accelerated but never spiralled; indexation did not return as an institution; and inflation decelerated without the mass-unemployment purge the 1970s playbook implied was necessary. The anchor, still intact from the credibility purchased decades earlier, did the heavy lifting: it denied the shock the second-round machinery, and the shock died. Two episodes, similar shocks, opposite outcomes — and the differentiating variable was not the inflation the economies experienced but the expectations structure the inflation landed on.
The strongest case against expectations
A serious strand of central-bank research argues that the expectations framework is built on sand: household survey responses are noisy, systematically biased upward, and driven by fuel prices; most price setters, when studied directly, do not think about aggregate inflation at all when setting their own prices; and the historical evidence that expectations cause inflation — rather than merely tracking it — is thinner than the theory's centrality suggests. On this critique, the expectations channel is a story macroeconomists tell because their models need it, and an investor who watches expectations data is watching an artefact of the modelling convention rather than a force in the economy.
The critique lands against a specific target — the use of measured expectations as a precise forecasting input — and largely misses the framework that matters for a portfolio. The anchor premium does not require households to hold accurate numerical forecasts or firms to consult surveys before setting prices. It requires only the behavioural fact the 1970s demonstrated in one direction and the post-pandemic episode in the other: that there is a regime in which nominal behaviour ignores realised inflation, and a regime in which it indexes to realised inflation, and the two produce entirely different shock dynamics. Whether one calls the difference "expectations" or "norms" or "credibility" is vocabulary; the regime distinction is observable in behaviour — the prevalence of indexation clauses, the frequency of repricing, the wage round's sensitivity to last year's headline rate. The investor's version of the critique is therefore a refinement, not a rejection: trust behaviour over stated expectations, treat survey levels as biased and survey changes as informative, and reserve the strongest weight for the long-horizon market measures and the behavioural evidence, which is where the anchor either holds or visibly slips.
What this changes for a portfolio
The operating conclusion is a hierarchy of attention that inverts the market's. The market's attention scales with the size of the actual-inflation surprise; the framework says attention should scale with the threat to the anchor. A hot print that long-horizon measures shrug off is consumer arithmetic and base-effect fodder — noise for the discount rate, whatever it does to the front end. A modest drift in distant-forward expectations, or the behavioural early-warnings — indexation demands appearing in wage negotiations, repricing frequency rising across firms — is the opposite: a small number that prices a regime. Sterling's Embedded Intelligence weights its inflation dashboard accordingly: actual prints route to the cash-flow and near-policy views, while the anchor gauges — long-horizon forwards, long-run survey changes, and the behavioural indicators — carry the discount-rate and regime weight. Sterling's rule of thumb compresses the whole piece: trade the temperature, but underwrite the thermostat — because every long-duration position in the book is, knowingly or not, a seller of insurance against the anchor slipping.
How to apply this framework
- Weight the long end of every expectations measure over the short end. Short-horizon surveys and breakevens mostly track fuel and headline conditions — a thermometer mislabelled as a thermostat. Distant-forward market measures and long-run survey responses are the anchoring gauges.
- Trust behaviour over stated expectations. Indexation clauses reappearing, repricing frequency rising, wage rounds keying off last year's realised inflation — these are the anchor slipping in the data that matters, whatever the surveys say.
- Read breakeven moves as three components, not one. Expected inflation, inflation risk premium, and liquidity all move breakevens. A widening driven by risk premium is itself information — the market charging more for anchor uncertainty — but it is not a forecast revision.
- Price central-bank forcefulness on expectations as anchor defence, not hawkish excess. The reaction function is convex in expectations: aggressive response to small expectation moves is the institution buying cheap insurance against an expensive repair — historically a supportive signal for duration once credibility is re-established, not a threat to it.